Under the hood, shinyobjects parses your code and moves around the arguments of reactive and rendered objects so that you can interact with them from the global environment. While not required, this workflow works best with a static input list that shinyobjects can help create. You can read more about the benefits of creating a dummy input list here
The main function, load_reactive_objects() takes the following steps:
parse the code that is active in the source pane or otherwise specified
.Rmd files, apps with a server.R file, or files using shinyServer(), runApp(), or shinyApp()
keep library(), load(), and assignment calls (<- or =), everything else is discarded
rewrite the expressions:
for reactive(x = ...), the x argument is moved to the body of a function. For eventReactive(event, valueExpr = ...) it uses the valueExpr argument. Thus
reactive_df <- reactive({ mtcars %>% head(input$n) })
becomes
reactive_df <- function() { mtcars %>% head(input$n) }
Now that it is a function, you can view and manipulate reactive_df()
for reactiveValues()
x <- reactiveValues(n = input$n)
becomes
x <- list(n = input$n)
for output assignments
output$x <- renderPlot(expr)
becomes
output$x <- recordPlot(expr)
for all other output assignments, the expr arguments are directly assigned
output$text <- renderText(paste("there are", input$n, "observations))
becomes
output$text <- (paste("there are", input$n, "observations))
evaluate the modified expressions into the specified environment, usually, the global environment
load_reactive_objects() has options to restart your R session and/or clear out your environment
There are two other functions:
convert_selection() does a shorter version of these steps using the code highlighted in the source pane view_ui() works with UI components. It takes either the html output in the console or selected code in the source pane and runs a shiny app of just that object.